The elephant in the room is that with the exception of Ethereum, all GPU-mined coins can relatively easily be "51%-attacked" by a small fraction of GPU miners.
There are ~10 million GPUs mining cryptocurrencies in the world today. Because the majority of them mine Ethereum (5-10 million are needed to generate 250 TH/s), it means the other coins can easily be overpowered and attacked if a small fraction of GPU miners decided to do so. For example look at Monero: 400 MH/s of CryptoNightV7 hashrate means there are 400k-800k GPUs behind it, therefore only 4-8% of the pre-existing worldwide GPU mining capacity is needed to attack it.
In a way we could say Monero is vulnerable to a "4-8% attack!"
The best way to defend a coin from this scenario is for it to implement a unique PoW algorithm that is very ASIC-friendly so that GPU miners couldn't overpower it. Of course the same attack scenario would exist if there is a large pre-existing installed based of this ASIC, so the PoW algorithm must be unique. For example Bitcoin Cash could currently be attacked by 10% of Bitcoin miners as they are both mined by the same SHA-256 ASICs.
The irony is that the misguided trend to try to be "ASIC-resistant" is actually worsening the value proposition of all these GPU-mined coins as it makes them more vulnerable to the very real possibility of 51% attacks...
Having a unique ASIC algo would raise the problem of the traditional 51% attack though.
Whichever fab produced it would have a huge hardware advantage in mining. They would mine themselves and just use 51% of their physical units (versus renting on a time-slot market) to do the attack.
...which kills the whole point of the attack. you already poured hundreds of thousands of dollars into fabbing those ASICs, which only work for that specific cryptocurrency. why would you do a 50% attack and tank its price? the only reason 50% attacks like on BTG is viable is that you can rent hashpower (no investment!).
Price crashes do seem to be a disincentive for a 51% attack for owned miners. How much would the price tank versus the gain possible with an attack? The two don't seem interlinked strongly.
The price-tank disincentive is multiplied by the block reward size -- the lower the reward, the the less the miner would care. The price-tank disincentive also is attenuated by future equilibrium mining. If you make the first ASIC and don't attack, future manufacturers will pop up as mining becomes more profitable. If you tank the price, mining profits may become low, but you thwart competition.
Attack gains are not connected strongly to the above disincentives. Attack gains are higher the more often you can cycle the coin -- you can actually steal way more than the market cap of the coin theoretically by cycling exchanges / other mechanisms of payment. Then you start running your 51% attack and reverse hours, days, weeks of transactions.
I see your point that there is some disincentive for a 51% attack, but I'm not sure that's enough.
This point doesn't seem to make sense. If you have 25 potential attackers, then for there to be an actual 51% attack, more than half of them (13/25) would have to band together to actually reach 51%. Unless the assumption is that the 25 parties buy chips and sit on them indefinitely without contributing to the network hash rate.
I'd guess they mean "better to have one potential attacker because of the centralization that mining with a unique ASIC would tend to cause, than to have 25 potential attackers because you're vulnerable to an attack by 1/25 of worldwide GPU power (which is normally used for other coins but could be briefly repurposed)".
There are ~10 million GPUs mining cryptocurrencies in the world today. Because the majority of them mine Ethereum (5-10 million are needed to generate 250 TH/s), it means the other coins can easily be overpowered and attacked if a small fraction of GPU miners decided to do so. For example look at Monero: 400 MH/s of CryptoNightV7 hashrate means there are 400k-800k GPUs behind it, therefore only 4-8% of the pre-existing worldwide GPU mining capacity is needed to attack it.
In a way we could say Monero is vulnerable to a "4-8% attack!"
The best way to defend a coin from this scenario is for it to implement a unique PoW algorithm that is very ASIC-friendly so that GPU miners couldn't overpower it. Of course the same attack scenario would exist if there is a large pre-existing installed based of this ASIC, so the PoW algorithm must be unique. For example Bitcoin Cash could currently be attacked by 10% of Bitcoin miners as they are both mined by the same SHA-256 ASICs.
The irony is that the misguided trend to try to be "ASIC-resistant" is actually worsening the value proposition of all these GPU-mined coins as it makes them more vulnerable to the very real possibility of 51% attacks...